(in lingua inglese)

Questa tesi, che fa parte di una ricerca più ampia e di lungo periodo sull’ efficienza energetica, si occupa di modellistica dinamica e simulazione di ruote entalpiche, avendo come obiettivo primario (ma in linea di principio non unico) quello di impiegare tali dispositivi per migliorare l’efficienza delle unità di trattamento aria (UTA). A tale scopo, per prima cosa viene analizzata la teoria di funzionamento delle ruote entalpiche, avendo però presente che il particolare ambito della ricerca presentata richiede modelli dinamici per studi a livello di sistema che includono anche il sistema di controllo.

Articoli tecnico scientifici o articoli contenenti case history

Pubblicato

Estratto del testo

realizzazione di questo lavoro. In particolare ringrazio il Professor Alberto Leva per avermi dato la possibilità di

impegnarmi in questo progetto e per la sua disponibilità, e l''Ingegner Marco Bonvini

che è stato per me un punto di riferimento durante il lavoro e si è dimostrato sempre

presente e disponibile. Inoltre ringrazio la mia famiglia, i miei amici e i miei compagni di corso che hanno

condiviso con me parte di questa esperienza e mi hanno sempre sostenuta.

2.2.1 Desiccant material . . . . . . . . . . . . . . . . . . . . . . . . 15 2.2.2 Rotating speed . . . . . . . . . . . . . . . . . . . . . . . . . . 17 2.3 System Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3.1.1 Characteristic potentials method . . . . . . . . . . . . . . . . 23 3.1.2 Wheel motion representation . . . . . . . . . . . . . . . . . . . 23 3.2 Empirical models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4.2.1 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 4.2.2 Physical phenomena . . . . . . . . . . . . . . . . . . . . . . . 30 4.3 Wheel motion representation . . . . . . . . . . . . . . . . . . . . . . . 32 4.4 Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.4.1 Balance equations . . . . . . . . . . . . . . . . . . . . . . . . . 33 4.4.2 Desiccant flow equation . . . . . . . . . . . . . . . . . . . . . 35 i

7.4.1 Control systems behavior . . . . . . . . . . . . . . . . . . . . . 57 7.4.2 Energy consumption . . . . . . . . . . . . . . . . . . . . . . . 58 7.5 Control system performance . . . . . . . . . . . . . . . . . . . . . . . 60

7.5.1 Improvement of the standard AHU control . . . . . . . . . . . 61 7.5.2 AHU with DW: PID controller . . . . . . . . . . . . . . . . . 63

with dynamic modelling and simulation of desiccant wheels, having as primary (yet

in principle not exclusive) goal that of employing such devices to improve the effi-

ciency of air handling units. To this end, first the theory of operation of desiccant wheels is analyzed, having how-

ever in mind that the particular purpose of the presented research requires dynamic

models oriented to system-level studies, that will eventually come to comprise and

consider also the control system. In other words, a peculiarity of the model devel-

oped herein, thus a contribution of the presented work, is that the developed model

has to find a suitable compromise among several different - and sometimes conflict-

ing - objectives, such as accuracy, ease of parametrization with design/nominal data,

which incidentally calls for a first-principle approach, and numerical efficiency. The results obtained in this work can be briefly summarized as follows. ' A dynamic model for a desiccant wheel is proposed that overcomes some limi- tations of these available to date in the literature, and a discussion is reported

- based on a convenient brief review of related work - to justify the statement

above. ' A Modelica implementation of that model is presented and validated against data available in the literature. Also, the so obtained simulation model al-

lows to investigate some aspects of the wheel operation that are exquisitely

dynamic, thereby paving the way to the control-oriented studies sketched out

below. ' Some possibilities are envisioned for the use of desiccant wheels in air handling units, viewing the addition of the wheel - that quite intuitively is not a novelty

perse - from an essentially control-centric standpoint. This allows to sketch

out some guidelines, to be further investigated in future works, for an objective

assessment of the achievable energy efficiency improvements. 1

za energetica, si occupa di modellistica dinamica e simulazione di ruote entalpiche,

avendo come obiettivo primario (ma in linea di principio non unico) quello di im-

piegare tali dispositivi per migliorare l''efficienza delle unità di trattamento aria

(UTA). A tale scopo, per prima cosa viene analizzata la teoria di funzionamento delle ruote

entalpiche, avendo però presente che il particolare ambito della ricerca presentata

richiede modelli dinamici per studi a livello di sistema che includono anche il sistema

di controllo. In altre parole, una peculiarità del modello qui sviluppato, quindi

un contributo del lavoro presentato, è che tale modello deve trovare un adeguato

compromesso tra diversi - e talvolta contrastanti - obiettivi, quali la precisione, la

facilità di parametrizzazione utilizzando dati di progetto e/o nominali, il che richiede

un approccio basato su principi primi, e l''efficienza numerica. I risultati ottenuti in questo lavoro possono essere brevemente riassunti come segue. ' Viene proposto un modello dinamico per una ruota entalpica che permette di

superare alcuni limiti dei modelli disponibili fino ad oggi in letteratura, e viene

riportata una discussione - sulla base di una breve revisione del lavoro già esistente

- per giustificare la realizzazione di un nuovo modello. ' Una implementazione Modelica del modello proposto è presentata e validata con

i dati disponibili in letteratura. Il modello così ottenuto permette di approfondi-

re alcuni aspetti del funzionamento della ruota che sono squisitamente dinamici,

preparando così il terreno per studi orientati al controllo qui di seguito introdotti. ' Vengono discusse alcune possibilità per l''uso di ruote entalpiche in unità di trat-

tamento aria, considerando l''aggiunta della ruota - che intuitivamente non è una

novità di per sé - da un punto di vista essenzialmente di controllo. Questo permette

di delineare alcune linee guida, che saranno ulteriormente approfondite nei lavori

futuri, per una valutazione oggettiva dei miglioramenti ottenibili dal punto di vista

dell'' efficienza energetica. In particolare, nel primo capitolo viene introdotto il contesto in cui questo lavoro

si inserisce e vengono presentati in maniera sintetica i sistemi di condizionamento

dell''aria usati negli edifici e il loro funzionamento. Viene inoltre messo in evidenza il

possibile risparmio energetico dato dall''uso delle ruote entalpiche. Nel secondo capi-

tolo viene illustrato il funzionamento delle ruote entalpiche e sono presentate alcune

configurazioni di unità di trattamento aria che prevedono l''uso di tali elementi. 2 Il terzo capitolo è dedicato a una breve presentazione di alcuni modelli di ruote

entalpiche presenti in letteratura, in particolare vengono presi in considerazione al-

cuni modelli di primo principio implementati in linguaggio Modelica e alcuni modelli

empirici. Il quarto capitolo illustra il modello proposto in questo lavoro mettendo

in evidenza le principali scelte modellistiche che lo differenziano dai modelli finora

proposti. Particolare attenzione viene posta alla rappresentazione del movimento

della ruota che risulta essere un elemento particolarmente importante per l''aderen-

za del modello alla realtà fisica. La scelta della modellizzazione della velocità della

ruota mediante un ''flusso di materiale', ci permette inoltre di tenere conto in ma-

niera semplice ed intuitiva di fenomeni fisici altrimenti difficilmente rappresentabili.

Tale modello è validato attraverso dati sperimentali presenti in letteratura (quinto

capitolo) con i quali viene riscontrata una buona aderenza. Nel sesto capitolo vengono presentati alcuni risultati di simulazioni per mettere

in evidenza la bontà del modello. In particolare, viene illustrata la distribuzione

spaziale dei parametri di interesse (temperatura e umidità) e viene inoltre riportata

un'' analisi parametrica atta a identificare la variazione del comportamento della

ruota al variare delle condizioni di funzionamento. Il tema del risparmio energetico è trattato nello specifico nel settimo capitolo, in cui

vengono messi a punto i sistemi di controllo sia per l'' UTA standard che per quella

con ruota entalpica in modo da poter confrontare la domanda energetica delle due

configurazioni nelle stesse condizioni operative. Nell''ottavo capitolo vengono illu-

strate alcune possibilità di controllo alternative che prevedono l''uso della velocità

della ruota come variabile di controllo. Nello stesso capitolo viene anche propo-

sta una strategia di controllo di temperatura basata sull''utilizzo, come variabile

manipolabile, della velocità della ruota. Nel nono capitolo sono presentati alcuni risultati di simulazione che illustrano i

consumi energetici dell''UTA in funzione della velocità della ruota in varie condizioni

operative. Viene quindi proposto un metodo per la scelta della velocità della ruota

in modo da minimizzare il consumo energetico dell''UTA, sia nel caso generale che in

quello in cui si possa usufruire di fonti di calore di bassa qualità e/o a basso costo,

come per esempio quelle legate a recuperi di calore ''di scarto' oppure alle sorgenti

rinnovabili. L''ultimo capitolo è dedicato alle conclusioni e agli sviluppi futuri. 3

and Air Conditioning) systems context, and to show the possible energy saving that

can be obtained by using the desiccant wheel (DW) technology. In particular, this

work is a part of a long-term research dealing with HVAC systems in buildings. Thus,

a system level view on HVAC systems is enough to understand the motivation of

our work. Energy saving is a crucial issue in the addressed context, and thus the

use of new technologies to improve it, is increasing in interest. Desiccant wheel

elements are one of the most promising technologies dealing with energy saving in

HVAC systems. Thus, the need of a model of such an element, that is suitable for

system design and diagnosis, appears clear. In the last part of this chapter, some

simulations are shown to prove the possible energy saving with DWs.

beings, a suitable operating environment for machines such as computers, or any

combination thereof. However, particularly when human comfort plays a relevant role, these conditions

reveal a somehow subjective character, as they also depend on each single person''s

feelings, which in turn may be time-varying. In any case, comfort depends mainly

on air temperature and humidity. Said quantities, thus need controlling, so as to

fulfill constraints expressed in terms of the so-called PMV (predicted mean vote)

index. The most typical constraint takes the form of (1.1): '' 0.5 ' P M V ' 0.5 (1.1) The PMV index is defined by the ISO 7730 standard, and the comfort is maximum

if its value is 0. If PMV is less than 0, a person tends to perceive the ambient as

''cold', while if it is greater than 0, a person perceives the ambient as ''warm'. Hence,

a comfort air conditions area can be defined as it is shown on the Psychrometric

Chart in Fig. 1.1. 4 1.2 Conventional Air Handling Unit

particular case at hand, is a very important phase for the success of a project. More specifically, when distinguishing types of HVAC systems, very important com-

ponents from the viewpoint of both comfort and energy efficiency are the AHUs (Air

Handling Units), on which we mainly concentrate.

AHUs, i.e., in our context, of AHUs not comprising a DW. Such units are composed 5 1.2 Conventional Air Handling Unit of several elements, as shown in Fig. 1.2. The main elements are: 1. Supply duct 2. Fan compartment 3. Vibration isolator (''flex joint') 4. Heating and/or cooling coil 5. Filter compartment 6. Mixed (recirculated + outside) air duct As shown in Fig. 1.3 in conventional AHUs (as opposite to twin coil ones described

rater on) there exists only one cooling coil. This coil is devoted to heat exchange

between chilled water and the heated air. The aim of this element is to remove

sensible and latent heat from the air. Thus, the chilled water used for this purpose,

has to be brought at a quite low temperature (typically around 7°C ). It should be noted that such a low temperature is required only for dehumidification.

In fact, to remove moisture, the air has to be brought at a temperature lower than

its dew point one. Thus, the chilled water temperature has to be considerably

low. Conversely, to obtain the desired air temperature without condensation, a

higher-temperature (around 15°C) chilled water would be enough. Therefore, a

conventional AHU deals with very low chilled water temperature because of the

condensation process. Clearly, after passing through the cooling coil, the air is at

an excessively low temperature, and it has to be heated before entering the room. In order to reduce energy consumption, due to the cooling of the water and the need

to heat the cold air after the condensation process, there are several solutions that

can be used. One of such solutions consists of the use of a twin coil AHU, which is presented in

the next section, but, as will be shown later on, the use of a desiccant wheel brings

advantages also for the problem just mentioned. 6 1.3 Twin Coil AHU

to the single coil (or conventional) one, the presence of a second cooling coil allows

to use each coil with different purposes. One coil cools and dehumidifies the fresh

air, using lower temperature chilled water, while the other coil is used to cool the

room air, using higher temperature chilled water. As such, the most of the chilled

water used here is at higher temperature with respect to that required by a conven-

tional AHU. Thus, with the twin coil configuration we can obtain a smaller energy

consumption according to the refrigeration cycle principle. 7 1.4 AHU with desiccant wheel

compression, condensation, expansion and evaporation. The energy performance of

such a cycle is typically quantified by the COP (Coefficient of Performance), defined

as: COP = QL Wnet,in (1.2) where QL is the cooling effect and Wnet,in is work input. The amount of energy consumed in the process is approximately proportional to the

difference between the evaporation and condensation pressures. It is also propor-

tional to the temperature difference between the same two cycle phases. Thus, the

energy consumption becomes smaller if the temperature difference between evap-

oration and condensation decreases. The condensation temperature is related to

the cooling water temperature, while the evaporation temperature is related to the

chilled water temperature. This implies that the higher the chilled water tempera-

ture is, and the lower the cooling water temperature is, then the smaller the energy

consumption is.

discussed more in detail in chapter 2. In this chapter, we only want to synthetically

justify the increasing interest in the DW''s technology, and underline the considerable

energy saving that can be obtain with the use of a DW. To this end, we present in

the next section, some simulation results from [13].

using a DW, the performance of a standard AHU and that of a AHU with a DW are

compared in reasonably realistic and operating conditions. The considered scenario

is illustrated in [13]: the AHU is used for the air conditioning of an office (volume

of 300 m3: 10 ' 10 ' 30m). The external ambient and indoor air conditions can be viewed in this context as

disturbances, which the control system has to reject. One of the most important

specifications on the HVAC systems deals with the number of the air renewal per

hour. In the presented configuration the AHU has to guarantee 4 air renewal per

hour, thus, only one third of the return air has to be recirculated. The aim of

the control system (working from 6:00 AM to 9:00 PM) is to maintain the air

temperature at 26° C, and satisfy a variable requirement for the absolute humidity (

between 12 '' 14gH2O/kgDryAir). The external air temperature and humidity profiles

are shown in Fig. 1.6 and Fig. 1.7, respectively the same figures also report the

temperature and humidity set points. The air recirculated from the office has a

temperature of 27° C, and its absolute humidity is 15gH2O/kgDryAir. 9 1.5 System energy performance

and right (red) boxes respectively represent the humidity and temperature control

systems. The air from the external ambient is first cooled and dehumidified by a coil (with

chilled water at 4° C). The absolute humidity is controlled by acting on the mass flow

rate of chilled water used in that coil. The air, once it exits the cooler, has the desired

moisture amount (thanks to the humidity control just mentioned), and its water

vapor content remains the same while the air subsequently passes through a rotary

heat exchanger. Then the air is heated with a second coil and, in the considered case,

a temperature controller regulates the mass flow rate of hot water at 45° C. As shown

before, humidity and temperature control can be treated separately (although they

need addressing in the correct order), as the controlled system is triangular: after

been dehumidified, the air is only heated until entering the room. In the presented

configuration two PI controllers have been used, but is clear that humidity control

affects temperature control. 11 1.5 System energy performance

and than enters the DW, where it is dehumidified and heated. Downstream the

wheel, the air passes through a heat exchanger, and then, it is post-cooled/heated

depending on the desired conditions. A second airflow is necessary to regenerate the

wheel, as will be discussed in sec. 2.2. This regeneration air, coming from the internal

ambient, first passes through the heat exchanger, and then is heated before entering

the wheel. The L-shaped (yellow) box identifies the humidity control system, which

regulates both the process and regeneration air temperature. This is because the

desorption capacity of the wheel depends on both temperatures. The right (red)

box identifies the air temperature control. The wheel velocity is considered as a

parameter, and set to 3rev/hour . The AHU with DW configuration will be discussed

more in detail in the following chapters.

good results for set point tracking, as shown in [13]. Unfortunately, however, the second layout has some small drawbacks: ' the system is higher in complexity; ' the AHU closed-loop response is slower with respect to the previous case, due to the inertia introduced by the wheel. 12 1.5 System energy performance Nevertheless, the important advantage of the DW configuration appears evident if

we consider the energy consumption. In the standard AHU, the air dehumidification

is performed through condensation, and to this end, the air temperature has to be

pulled down considerably. Thus, the standard AHU is quite energy demanding

because both the cooling energy, used by the condenser, and the heating energy,

used to heat the air after dehumidification, are relevant. The DW configuration

can reduce such a problem, because the dehumidification process is performed by

the wheel through the desiccant material adsorption process. To maximize the advantages, as will be shown more in detail in the following chapters, the rationale

is to cool down the process air and heat the regeneration one.

consumptions of the two configurations is compared. It is clear that DW configu-

ration reduces the total energy consumption, and in particular the cooling energy

demand. Results, listed in Tab. 1.1, confirm that the total energy consumption can

be reduced up to 37%. Energy consumptions standard AHU AHU with DW Cooling energy [kWh] 116.94 72.1 Heating energy [kWh] 30.34 21.8 Total energy [kWh] 147.28 93.9

interest for energy saving purposes in HVAC systems. In this chapter, the physical

structure and the operation of a DW are presented. Furthermore, some desiccant

air conditioning system configurations are illustrated, to show how DWs can be used

in the context of air conditioning.

regulation process is based on a cooling coil and a heating one: the air is first cooled

below its dew point temperature, and then heated until the desired temperature is

reached. This overall process can however be realized through more efficient HVAC

systems based on the use of DWs. A desiccant wheel is composed by a large number of channels, each one made by

a supporting material (matrix) coated with desiccant. The wheel is also angularly

partitioned by a clapboard in two different areas, the process area and the regenera-

tion one, and driven by a motor which imposes a desired angular velocity, as shown

in Fig. 2.1. 14 2.2 Working behavior of the wheel

directions: these are the process airflow and the regeneration one, as shown in

Fig. 2.1. Process air, humid and at ambient temperature, flows through the so-called process

area, where a part of the water vapor contained in it, is adsorbed by the desiccant

material of the wheel. Regeneration air flows through a usually smaller (or at most equal) area, called the

regeneration area. Before the regeneration air passes through the wheel, it is heated

up. During the regeneration process, water contained in the wheel is extracted from

the desiccant by the airflow, and the desiccant is regenerated. The wheel rotation brings the desiccant material alternatively in the process area

and in the regeneration one. Passing through the regeneration area, the desiccant

material is brought back to the condition it had when last entering the process area,

and the adsorption/desorption cycle can start again. The adsorption capacity is thus intuitively foreseen to be a function of desiccant

material, angular speed of the wheel, process and regeneration areas ratio, geometry

of the wheel, and of course temperature, humidity and velocity of the airflows. Therefore, the choice of the desiccant material plays a crucial role in the design of

the wheel, and significantly affects the performance of the whole air conditioning

system.

are however some, the so-called desiccant materials, in which said capability is par-

ticularly relevant; among these are e.g. activated carbon, activated alumina, silica

gel, lithium chloride, and calcium chloride. The most commonly used adsorbent for

DWs is silica gel, i.e., a porous, amorphous form of silica (SiO2). Silica gel has a

great affinity for water vapor due to the enormous quantity of microscopic pores:

the internal surface area of pores is in fact several orders of magnitude larger than

the outer surface area of the adsorbent. Adsorption of water vapor involves two different processes: chemical sorption and

physical adsorption. The first process is permanent, and cannot be reverted by

regenerating the silica gel, while the physical adsorption is a reversible process,

driven by the intermolecular forces of attraction called Van Der Wall forces, that

hold water molecules on the pores surface. Thus, it is the physical adsorption that

plays a crucial role in dehumidification process involving DWs. 15 2.2 Working behavior of the wheel An important characteristic of a desiccant material is its adsorption isotherm, which

determines its water vapor adsorption capacity as a function of temperature and

vapor pressure. We can classify a desiccant material according to the average diameter of the pores,

but also, and more interesting for our purpose, according to its adsorption isotherm. Some types of adsorption isotherms are shown below in Fig. 2.2, where the nor-

malized loading fraction (NLF: actual desiccant water content at corresponding

RH/maximum desiccant water content at RH=100% ) is a function of relative hu-

midity (RH), defined as the vapor pressure in moist air over the vapor saturation

pressure at the same temperature, i.e., RH = Pv/Pvs(t).

prescribed temperature. In this section, we consider adsorption isotherms to analyze

the adsorption capacity of the corresponding material. It is clear that curves with

great slope at low RH, e.g. 1E, correspond to materials that have a great adsorption

power at low relative humidity, but saturated faster, i.e. their adsorption power

increases ''fast' and then ''slower' with the increase of RH. For this type of materials,

as shown in Fig. 2.2, the NLF reaches almost the maximum value at low RH (e.g.

RH ' 0.3 for type 1E isotherm refers to N LF '0.98). On the contrary, materials

with e.g. type 3E isotherm, do not have much adsorption power at low values of

RH, but they express all their adsorption capacity when dealing with high RH (e.g.

for type 3E, for RH ' 0.8 the adsorption capacity increases ''faster' with the RH). Quite intuitively, thus, in the DW context it is also necessary a desiccant material

should possess large saturated adsorption amount, but it is also necessary to take

into account that this material has to be easily reactivated during the regeneration

precess. The energy performance of a DW depends mainly on the power used for

the regeneration process. Thus, the desorption process of the desiccant has to be 16 2.2 Working behavior of the wheel considered. Referring to Fig. 2.2, type 1E materials, because of its nearly complete

loading at low RH, are more difficult to regenerate. Thus the adsorption performance

of the desiccant material should approach type 1M adsorption isotherm.

shown. This process exhibits an hysteresis, as the adsorption behavior of the mate-

rial is different from its desorption behavior. In other words, the amount of energy

that the material needs to be reactivated, depends on the desorption process. Thus,

it is clear that the adsorption and desorption processes - rigorously speaking - re-

quire a different amount of energy. However, in this work we neglect the hysteretic

behavior of the desiccant material owing to its tendentiously modest entity. As said before, the choice of the desiccant material is very important, and thus, there

are several researches under way to find the desiccant material that best approaches

type 1M in its adsorption performance. At present, commonly used desiccant ma-

terials exhibit some drawbacks. For example, the silica gel adsorption performance

decreases quickly with the rise of temperature, while lithium chloride has a higher

adsorption capacity, but it also shows the lyolysis phenomenon, which leads to the

loss of desiccant material and may reduce the performance of the entire system.

process and regeneration air inlet conditions, process air outlet humidity depends on 17 2.3 System Configurations the wheel velocity, as shown in Fig. 2.4. We can identify two main different working

behaviors of the wheel, depending on its velocity. First, it is clear that at low speed the desiccant material remains for a long time in

the process area, and its dehumidification capacity is exhausted before the material

comes into contact with the regeneration air. Increasing the rotating speed, the ad-

sorption capacity is better exploited, and water content in outlet process air reaches

a minimum. If however we continue to further increase the speed, desiccant material

comes to not use all its adsorption capacity, because the time spent in the process

area is too short with respect to the adsorption time constant. In such a case, the

process is dominated by heat transfer, and the dehumidification rate decreases. In Fig. 2.4, we show an example of such a behavior. Note that there is an optimal

rotating speed for dehumidification, corresponding at the minimum water content in

the outlet process air. For different operating conditions the optimal wheel velocity

can be found and exploited to maximize the wheel efficiency.

mance of the whole air conditioning system, the system configuration is another

crucial issue. In this section, we synthetically present some commonly used con-

figurations, and we underline some advantages and drawbacks of them. For this

purpose, it seems interesting to show, for each system configuration, both the plant

scheme and the physical transformations undergone by air, using the Psychrometric

chart. It should be noted that the dehumidification process in the DW is close to a

isenthalpic transformation, as shown in Fig. 2.5 (1-2). So, additional air conditioning

equipments must be introduced in order to reach the desired air conditions. 18 2.3 System Configurations

nington in 1955 [14]. In Fig. 2.5 the Pennington cycle, also known as ventilation

cycle, is presented. The ambient air, at state point 1, is directly used as process

air for the DW. The air, passing through the DW, decreases its moisture content

and increases its temperature, because of the adsorption heat effect, until it reaches

the state point 2. Then, this hot dry air is sensibly cooled, from state point 2 to

3, in a heat exchanger (HE). The process air is evaporatively cooled, from 3 to 4,

by passing through a direct evaporative cooler (DEC). Then, the air can enter the

room in the desired conditions, represented by state point 4. On the regeneration

air side, return air, which is taken from the inlet ambient at state point 5, is cooled

and humidified in another DEC, from state point 5 to 6. This air exchanges sensible

heat with the process air, from 6 to 7, to precool the process air and pre-heat itself.

Then the regeneration airflow is heated, from state point 7 to 8, by the heat source

(HS). This air, at state point 8, can finally be used to regenerate the DW. After the

regeneration, the air is exhausted and can be released at state point 9. There are some variations of Pennington cycle, introduced in order to improve the

performance of the system and to make it more suitable to the desired operating

conditions. For example, the Recirculation cycle (Fig. 2.6) reuses return air as pro-

cess air, to increase the cooling capacity, as the internal ambient air is at humidity

and temperature conditions that are closer to the desired ones, if compared to the 19 2.3 System Configurations external ambient air. This fact is evident, if one compares stare point 1 in Fig. 2.5

and the same point in Fig. 2.6. Thus, with a recirculation cycle, the variation of the

air conditions requires less energy consumption. In Recirculation cycle the ambient

air is used for regeneration purpose with the same advantages shown for the process

airflow. The main disadvantage of this cycle is however some lack in fresh air. In

HVAC systems the renewal of internal air with external ambient air is very impor-

tant for human health, as already noted, and there are several regulations in this

regard.

ing systems is the staged regeneration. The basic idea was proposed by Glav in

1966, [8], and then rearranged in the presented context for desiccant regeneration.

As illustrated in Fig. 2.7, the particularity of the Staged Regeneration Cycle, lies

in the regeneration side of the DW, which is divided in pre-regeneration side and

regeneration side. After being pre-heated in the HE, from state point 6 to 7, only a

fraction of the regeneration air is heated by the HS, from state point 7 to 8, while

most of the regeneration air is directly introduced to the pre-regeneration area of the

DW. The advantage of this configuration is that the desiccant is first pre-regenerated

with a low temperature airflow, and then is further regenerated with a much smaller

amount of high temperature heat. This configuration thereby allows less energy

consumption for the regeneration process. 20 2.3 System Configurations

best of them depends on the particular application context we do not delve into.

Further details, however, as doing so would substantially stray from the scope of

this dissertation. 21

a satisfactory behavior of the whole system. Thus, a mathematical model for such

a component can be extremely useful in system design, but also in analyzing exper-

imental results, and all in all to find the optimal wheel operation for the envisioned

working conditions. In the literature, some DW models are presented and classified, as shown in [16].

Mathematical models can be divided into two main categories: first principle models

and empirical models.

ena involved in the wheel operation. They can have a different level of accuracy,

depending on how said phenomena are represented. In particular, sticking to the

purpose of this work, if we want to investigate the operation of the wheel for a

system level study, a very accurate description is not required. The physical phenomena involved in the wheel behavior deal with heat and mass

transfer principles, but some of them are of non trivial modelling at all. Thus,

in order to obtain a simple enough model, some ideality assumptions are in order

and advisable for example, some factors that do not significantly affect the wheel

behavior, as seen in the context of the complete AHU, can be often neglected, thus

reducing the complexity of the model. First principle mathematical models are composed by partial differential equations,

based on the mass and energy balances, and involving mass and heat transfers. In

addiction, the initial conditions of airflows and desiccant material and the boundary

conditions have to be determined so as to close the model. As said in sec. 2.2.1,

the adsorption/desorption process is the key element on which the wheel operation

is based. Thus, this process has to be represented with a specific equation. How

to model this phenomenon is not a trivial matter, however in the literature, some

semi-empirical models are presented and used to predict the adsorption capacity of

the material, e.g. [5]. There are several methods to solve the differential equations used to describe the

physical phenomena. One of them is the characteristic potentials method. 22 3.1 First principle models

non-linear partial differential equations into a set of decoupled differential equa-

tions. For this purpose, new independent variables (which are called characteristic

potentials, whence the method''s name), are used. These are associated with new

parameters, which can in turn be related with convenient specific-heat ratios. The

new set of equations is analogous to the heat transfer equations in a rotary heat

exchanger. Thus, the problem con be solved with the analogy method, drawing e.g.

from the solution of Kays and London [18] for a rotary heat exchanger. This method has some relevant advantages. First, from the analogy with a ro-

tary heat exchanger follows a simplified solution, requiring a low computation time.

Then, the performance prediction is based on parameters, and relationships similar

to those used for heat exchangers, and the model operation can be easily understood

with the use of psychrometric charts. However, the parametrization of the model is not a trivial matter. To partially

overcome this problem, the method can be simplified, e.g. one can assume the

effectiveness of the two characteristic potentials as constant, and the potentials can

be described by analytical formulae (for silica-gel, for example, those defined by

Jurinak [11]). The most relevant drawback of this (further) simplified model, is that

default values for the effectiveness should be used, as they are generally unknown.

Since no model completely fulfils the requirements defined above, a new model

has been developed based on the method of characteristics. The model has been

adapted in order to be parametrized only with the data provided by desiccant wheel

manufacturers.

motion. There are several works dealing on this problem. The aim of this subsec-

tion is to present three different modelling approaches, based on Object Oriented

Modelling (OOM), i.e. the context in which the present work takes place. In our work, we decided to use the Finite Volumes approach, that is a suitable

method for the model implementation using the available software paradigm (the

Modelica language, applied with the Dymola translator). The Finite Volumes approach is based on control volumes (CVs) for which balance

equations are written. In DW modelling, said volumes represent wheel ''pie slices'

and they are alternatively traversed by the process and regeneration airflows, de-

pending on their position during the DW rotation. However, Modelica can only

handle ordinary differential equations with respect to time, however, and to over-

come this problem, there are some different approaches. 23 3.1 First principle models In Casas et all. ([17]), a variable transformation is used to express the position of

the volume in terms of time instead of angular position. This is possible under the

assumptions that the velocity of the wheel is constant and the boundary conditions

do not change during the differential model''s integration time. The wheel is split

in two shares, one for each airflow, as shown in Fig. 3.1. The referring volume is

composed by two opposite pie slices, one in part I and the other in part II. At the

end of half a rotation period, the boundary conditions of the two slices are switched.

but owing to the abruptly described (switching) slice transition just mentioned,

the model ends up producing discrete output variables even from continuous input

values. Second, the model computing time is quite large, because (referring to the

typical solvers used in OOM tools, that exploit variable-step integration to improve

efficiency) it causes a state event every half a revolution time. To overcome said

problems, another approach is presented in [1]. As shown in Fig. 3.2, the wheel

motion is represented by the use of a so called desiccant flow, which flows in angular

direction. Control volumes are spatially fixed (i.e., the model refers to portions of

the fixed ambient space and not of the wheel moving in it) and composed by an Air

CV and a Desiccant CV. Said volumes interact by exchanging heat and moisture

for modelling the adsorption/desorption process. The DW velocity is thus related

to the desiccant mass flow rate through the Desiccant CV. 24 3.1 First principle models

valves to represent the wheel motion. Each pie slice of the wheel is fixed in space,

and alternatively traversed by the two airflows depending on the wheel motion. As

shown in Fig. 3.3, the airflow (non abrupt) switching is represented by the opening

and closure of four valves, and the time at which valves change state is related on

the wheel rotation speed.

elements that make the representation of the wheel motion very different from the

physical phenomenon. Also, due to the periodical behavior of the model, the draw-

backs already mentioned for the Casas et all. model ([17]) appear again. 25 3.2 Empirical models

complex for a system analysis, because solving differential equations is complicated

and very time-consuming. A different approach was thus followed by Beccali et

all. in [12], in which an empirical model for evaluating the performance of a DW

was proposed. Using a set of equations, based on second-order polynomials, they

provided outlet air temperature and humidity as a function of the inlet regeneration

and process airflows. Using the available experimental data, they could define a

set of parameters (with no direct physical meaning), however which could be used

to predict the performance of the wheel. The simulation results obtained with this

model are very close to the experimental data, but obviously, if the physical system

changes, parameters have to be recalculated, and there is no way to simulate a wheel

that thus not yet exist, as it is being designed and there is obviously no experimental

data on it. Further, other works provide simplification for the model presented in

[12], devoted mainly to reduce the number of parameters. 26

search includes design, we can further (and in some sense alternatively with respect

to the view point of chapter 3) divide the DW literature works in two main cate-

gories: ' complex ones, that describe DW in a very detailed manner, but are compu- tationally intensive and time demanding, thus not suitable for a study that

requires a huge number of simulations; ' simple ones, that are longer usable, but natively created for steady-state stud- ies and their parameters have no direct physical meaning. The aim of the present work is to propose an innovative model that has the main ad-

vantages, and overcomes the most relevant drawbacks, of both categories. Therefore,

the new DW model is based on first-principle equations (thus with physically mean-

ingful parameters) and includes spatial discretisation at a system level, as shown in

Fig. 4.1. The wheel has length L and radius R and it is split in M slices (radial direction) each

one divided in N control volumes (axial direction); each slice has a corresponding

angle θ = 360/M and a length dl = L/N. As shown in sec. 2.1, each slice is composed

by several channels. In this chapter, we first focus on the single channel, to understand in detail the

physical phenomena involved in DW operation. Then, we introduce the used control

volume, and the balance equations used in the presented model.

not take into account some phenomena which would make it too complex for our

purposes. Thus, before introducing the governing equations for the single channel,

it is required to underline the preliminary assumptions on the base of the model

construction. These assumptions are listed below. ' Axial heat conduction and water vapor diffusion in the air are neglected; ' hysteresis in the sorption isotherm for the desiccant, shown in sec. 2.2.1, is neglected, and the heat of sorption is constant; ' all channels are identical, with constant heat and mass transfer surface areas, adiabatic, and impermeable; ' the matrix thermal and moisture properties are constant, as are the mass and heat transfer coefficients, and the adsorption heat per unit mass of adsorbed

water; ' mixing between the process and the regeneration airflows is neglected.

Most of them aim at analyzing the possible geometries for the duct, to find advan-

tages and disadvantages that different choices of the channel characteristics yield; for

example in [6] the authors take into account three different channel shapes: square,

triangular and circular. In our case, the wheel channels have the shape shown in

Fig. 4.2 and can be described by two scalars, a and b, that define respectively the

height and the length of the channel. Note however that the proposed modelling ap-

proach is general, and according for other channel shapes is totally straightforward. 28 4.2 Single channel

effective area, Aeff , and the lateral area, Axcg. The effective area represents the wheel frontal area subtracted by the area occupied

by the support matrix, thus it is the frontal area through which the airflows actually

traverse the wheel, and is defined as: Aeff = ab '' ' b 2 + 2 a '' b 2 !! 2δ (4.1) where 2δ is the thickness of the matrix structure (with the desiccant). The lateral area represents the contact area between air and desiccant material that

is involved in the heat exchange and humidity adsorption/desorption processes, and

it is defined as: Axcg = 2 ' b 2 + 2 a '' b 2 !! dl (4.2) where dl is the channel element length. 29 4.2 Single channel

phenomena involved in the DW behavior. First, we define the water mass flow rate adsorbed by the desiccant material, positive

if adsorbed, as a function of the adsorption isotherm of the desiccant material, see

[5], as shown in sec. 2.2.1. wad = Md ' 0.24' 2

3 '' Xd (4.3) where M d is the mass of the desiccant material, ' is the air relative humidity, ' is

a discharge time constant and Xd is the desiccant water content. The time constant ' is defined as: ' = Md Axcghm (4.4) where hm is a mass transfer coefficient, while the desiccant water content Xd is

defined as: Xd = M wd M d (4.5) where M wd is the mass of water inside the desiccant material. Associated to such an adsorption/desorption process, there is a heat flow from the

air to the desiccant material due to thermal convection, i.e., qcam = htAxcg (Ta '' Tm) (4.6) where ht is the convective heat transfer coefficient and T a , T m are respectively the

air and matrix material temperature. In addition, there is a power associated to the water mass flow rate adsorbed/released

by the desiccant material, defined as: Qwam = wad hwv, hwm, wad ' 0

wad < 0 (4.7) where hwv is the water vapor specific enthalpy and hwm is the specific enthalpy of

water in the desiccant material. 30 4.2 Single channel Another phenomenon, that we need to take into account, is the water vapor diffusion

in the desiccant material: this it is particularly relevant at slow wheel velocity in

the angular direction, but has some relevance also in axial direction because of the

spatial discretisation used in the model. We define the water mass flow rate in the desiccant material between two different

volumes as: win2 = G (Xd1 '' Xd2) (4.8) where win2 is the water mass flow rate from the first to the second volume, Xd1 , Xd2

are respectively the desiccant water content in the first and in the second volume,

and G is a diffusive coefficient. The diffusive coefficient G is defined as: G = coef Ad dx (4.9) where coef is a diffusivity-related constant depending on the desiccant material and

Ad and dx are the crossing area and the distance between the two regions. These

parameters are defined, for a single channel, as: Ad = 2δddl (4.10) dx = ' b 2 + 2(a '' b 2 ) (4.11) where δd is the only desiccant thickness, if we deal with water vapor diffusion in

desiccant material in angular direction or Ad = Axcg dl 2δd (4.12) dx = dl (4.13) if the axial direction is being considered. 31 4.3 Wheel motion representation

the wheel rotation. In the literature there are several DW models that use different

modelling approach in this regard, as shown in sec. 3.1.2. To represent the wheel motion, we suppose that there is a desiccant flow that flows

through the wheel in angular direction, according to the wheel speed. Such a flow is

composed by the desiccant (with its water content) and the matrix material. Each

control volume is thus spatially fixed (i.e., it does not move with the wheel), and

it is traversed by either the process or the regeneration airfolw (depending on its

angular position) in axial direction , and by the desiccant flow in angular direction

(Fig. 4.3).

respect to which the mass and energy balances can be defined. Obviously, the

physical phenomena in the control volume are structurally the same as those of a

single channel, discussed in sec. 4.2.2, which is rigorous as long as one considers the

number of channel in a volume to be given by: n = ''R2 ab · 360' (4.14) where ' is the angle of the slice and R is the wheel radius.

control volume shown in Fig. 4.4. ' Mass balance of desiccant and matrix (solid) material: · Md = mdesin '' mdesout = 0 (4.15) where Md is the mass of desiccant material and mdesin, mdesout are respectively the

incoming and outgoing desiccant mass flow rates. 33 4.4 Equations ' Mass balance of moist air: · Ma = min '' mout '' wad (4.16) where Ma is the mass of moist air, min, mout are respectively the incoming and

outgoing moist air mass flow rates and wad is the water mass flow rate adsorbed by

the desiccant material and defined in 4.3. ' Mass balance of water in the desiccant material: · Mwd = mdesinXdesin''mdesoutXdesout+wad+wdl''in''wdl''out+wdθ''in''wd'''out (4.17) where Mwd is the mass of water vapor in desiccant material, mdesinand mdesout

are the incoming and outgoing desiccant material mass flow rates, while Xdesin ,

Xdesout are the water vapor mass fraction respectively associated to said flow rates,

wdl''in, wdl''out, wdθ''in, wd'''out are the incoming and outgoing water vapor mass

flow rates, in the dl and dθ directions, due to the diffusion in desiccant material and

defined in 4.8. ' Mass balance of water in the moist air: · Mwa = minXin '' moutXout '' wad (4.18) where Mwa is the mass of water vapor in the air and Xin, Xout are the water vapor

mass fraction associated to min and mout. ' Energy balance in the moist desiccant material: · Em = mdesinhdesin '' mdesouthdesout + wdl''inhwdl''in '' wdl''outhw+ (4.19) +wdθ''inhw '' wd'''outhwd'''out + Qwam + qcam where Em is the energy of the matrix and desiccant material, hdesin, hdesout are

the specific enthalpies associated to min and mout, Qwamis the power associated to

the water mass flow rate adsorbed/released by the desiccant material defined in

4.7, and qcam is the heat flow from the air to the desiccant material due to thermal

convection, and defined in 4.6. ' Energy balance in the moist air: · Ea = minhin '' mouthout '' Qwam '' qcam (4.20) where Ea is the energy of the moist air contained inside the volume. 34 4.4 Equations

flow in the angular direction. More in detail, said desiccant flow is represented with

the relation below: mdesout = R2 2 · dl · ' · (ρD + ρM ) ! · 1 + M wm (MD + MM ) ! ; (4.21) where ' is the angular velocity of the wheel [rad/sec], ρD and ρM are respectively

the desiccant material and the matrix densities, M wm is the water content in the

desiccant material, MD and MM are the mass of desiccant and matrix material in

the control volume. Equation 4.21 expresses the mass flow rate of the desiccant flow moving into the

wheel as two components: ' a constant part of solid desiccant material, R2 2 · dl · ' · (ρD + ρM ); ' a quantity depending on the water content in the solid desiccant, R2 2 · dl · ' · (ρD + ρM ) · M wm MD+MM . 35

in the literature [2, 3, 4, 5]. The desiccant wheel used for the comparison has the

characteristics listed below. ' desiccant material: silica gel; ' radius: R = 1m; ' thickness: L = 0.2m. ' channel dimensions: a = 3.2mm, b = 1.8mm; Referring to Fig. 5.1, the input data are the inlet air humidity of the process and

regeneration airflows (Xpro, in, Xreg, in), and the inlet air temperatures of the

same air streams (T pro, in and T reg, in). In the simulations, said inputs have been

set according to the experimental data, as shown in detail in the next section. In all

the case studies, the inlet humidity of the process air is equal to the inlet humidity

of the regeneration air, and the velocity of both air flows is equal to 1m/s.

We decided to use the same data to provide our model validation. Tab. 5.1 shows

the inputs used for the comparison between simulated and experimental data. 36 5.1 Comparison between simulation results and experimental data EXP/SIM Xpro/reg, in[g/kg] T pro, in[°C] T reg, in[°C] Apro

Areg [m 2/m2] expA 4.4-5.0 18.0 100 1 expB 13.2-13.8 25.0-27.0 60 1 expC 7.8 24.9 140 3.3 simA 4.4 18.0 100 1 simB 13.5 25.0 60 1 simC 7.8 24.9 140 3

state, against the wheel rotating speed and can be compared with Fig.5 in [7].

ticular, the effect of the wheel velocity on the outlet conditions of the process air,

is caught and reproduced, consistently with the discussion of sec. 2.2.2. At ''low' 37 5.1 Comparison between simulation results and experimental data rotation speed, any given portion of the wheel matrix remains in contact with both

the process and the regeneration air streams for a ''long' time. As a consequence,

the desiccant water vapor content has the time to reach the maximum value made

possible by the operating conditions, before the wheel portion carrying that desic-

cant, is led by rotation back to the regeneration area. Thus, there is a limit on how

much drying can be exerted on the process air. On the contrary, at ''high' rotation

speed, the desiccant does not remain ''long enough' in the process area, and it is

still able to adsorb water when it enters the regeneration area. Thus, in both the

''extreme' (very fast or slow rotation) conditions, the desiccant adsorption capacity

is not well exploited, and therefore it is clear that there has to exist an optimal

revolution speed, which minimizes the process air outlet humidity. Most important,

this optimal speed can be predicted by the presented model. Looking at the experiments a bit more in detail, it can be noted that we obtain

good agreement between model and data in particular for expA and expB, where

the maximum difference between simulation results and experimental data is less

than 10%, and the optimal revolution speed is well predicted. In the case of expC we conversely obtain slightly less precise results: in particular,

the maximum error between data and simulation results is less than 20% and the

optimal speed is not so well predicted. It is however worth noting that the expC

data appear to possibly contain some outliers, which can be at least hypothetically

detected by further looking at the particular operating conditions. The main dif-

ference between expC and the other experiments lies, in fact, in the process and

regeneration areas dimensions. As shown in Tab. 5.1, in expA and expB the wheel

is split by the clapboard in half, thus the process and regeneration areas are equal,

while in expC the process area is 270°. Given the geometry of the experimental

setup, it can be assumed that guaranteeing the same air flow velocity conditions

in both cases (as the model assumes) is not trivial. One can thus suppose that

the observed discrepancies are due to unmodelled variations in the air speed, and

since the entity of the observed errors is compatible with the magnitude that those

discrepancies can be supposed to have, still be convinced of the model validity. 38

ior of a DW unit, has been shown. This chapter is conversely devoted to using the

same model fore a more detailed analysis of the internal operation of the wheel. In

other words, the detailed modelling approach followed here has already shown capa-

ble of representing the wheel ''seen from outside'; the aim of this chapter is to prove

that such an approach is really representative also of the wheel ''seen from inside'.

To this end, we first define a base case, and show the spatial distribution of some

interesting quantities, e.g. temperature and humidity, referring to that case. Sub-

sequently, a parametric analysis is performed to underline how physical parameters

influence the wheel operation.

same DW used for validate the model in chapter 5. The base case has the features listed in Tab. 6.1. airflow T,in [°C] X,in [g/kg] area angle [°] velocity [m/s] process 30 15 180 1 regeneration 80 15 180 1

N=5 volumes per slice were used for the simulations. The wheel is divided in half,

maintaining the 6 upper slices (from 1 to 6) in the process area, and the lower 6

(from 7 to 12) in the regeneration area. Within each of the 12 slices, process air

enters the wheel in volume 1 and exits from volume 5, while regeneration air enters

in volume 5 and exits from volume 1. The wheel rotation is counterclockwise (see

the arrow in Fig. 6.1 for the rotation direction assumed as positive). To prove that the model provides a good representation of the DW behavior, the

distribution of air temperature (Ta), desiccant material temperature (Tm), water

content in the air (X) and in the desiccant (Xd) are shown in each volume (Fig. 6.2,

Fig. 6.3, Fig. 6.4, Fig. 6.5) for a rotating speed of 10 rev/h. 40 6.2 Spatial distribution analysis

shown. The two temperatures differ up to at most 4°C in the boundary slices, thus

between the process and the regeneration areas, and show the same trend. First, we analyze the process area. We start considering volume 1, that is directly connected with the inlet process air.

The desiccant material, which is heated by the regeneration air, has a high tem-

perature while entering the process area in slice 1. Process air stream temperature

is influenced by desiccant condition entering the process area, and thus it is higher

than the prescribed 30°C process temperature also in volume 1 of each slice. Both

the mentioned temperatures decrease from slice 1 to slice 6, because the air stream

cools down the desiccant material while said material flows in angular direction due

to rotation. If we conversely analyze the axial direction, from volume 1 to volume 5

of the same slice, the same two temperatures increases. Now we analyze the regeneration area. If we consider volume 5 instead of volume 1, we can see a dual effect with respect to

the process area just addressed. The desiccant material enters the regeneration area,

in slice 7, at a low temperature referring to the air flow regeneration temperature of

80°C. Then the material is heated by the regeneration air, and both the considered

temperatures increase from slice 6 to slice 12. In the axial direction, thus following

the regeneration air stream direction, both temperature conversely decrease.

shown, to prove the good agreement between the model and the assumption made

(and verified in chapter 5) on the wheel operation. If we consider Fig. 6.4, we can see that the air water content decreases from volume

1 to volume 5 for process area slices, while it increases from volume 5 to 1 for

the regeneration ones. Thus the process air is dehumidified, while the opposite

happens to the regeneration air. If we consider volume 5, thus the outlet process air

humidity, we can make the following statements. At the beginning of the adsorption

period, from slice 1, the adsorption process is less effective because the desiccant

material is still at high temperature, as said before and shown in Fig. 6.3. Then,

at the end of the adsorption period, the desiccant increases its water content, as

shown in Fig. 6.5, and its adsorption capacity decreases. Thus, the air humidity in

the process area (from slice 1 to 6 of volume 5) first decreases and then increases,

because the adsorption phenomenon is influenced by the desiccant water content

and temperature. At the beginning (slice1) the effect of the temperature on the

adsorption capacity is more relevant, while it is the desiccant water content at the

end (slice 6). We can see the dual effect considering volume 1 in the regeneration

area: from slice 7 to 12, air humidity first increases and then decreases. Referring to Fig. 6.5, we notice that water content in the desiccant material is higher

in volume 1 than in volume 5, in both the regeneration and the process area. This

is because of the airflows'' direction: process air releases water flowing from volume 43 6.3 Parametric analysis of performance 1 to volume 5, while regeneration air brings water flowing from volume 5 to volume

1. Thus the water content in desiccant is greater at the process air inlet and at the

regeneration air outlet, corresponding to volume 1.

trated in sec. 6.1. The wheel operation, as said in sec. 2.2, is influenced by some

parameters. For the presented analysis we decided to change, one at time, the inlet

conditions of airflows, and to show how these conditions affect the wheel behavior.

In Fig. 6.6 and Fig. 6.7 the simulation results for outlet process air absolute humidity

(Xpro,out), as a function of the wheel speed, are presented, and can be compared with

Fig.6 in [7].

temperature, velocity and absolute humidity with respect to the base case. ' If the inlet process air temperature decreases (blue line) the relative air hu- midity becomes higher at constant absolute humidity. Thus the desiccant

dehumidification capacity increases, and the process air exits the wheel with a

lower amount of water. This effect is due to the higher capacity of the desiccant

material of keeping water at high relative humidity as shown in sec. 2.2.1. 44 6.3 Parametric analysis of performance ' If the process air velocity decreases (green line), at low rotating speed, also the outlet air relative humidity does. This is because the process air remains

for a longer time inside the wheel. However, at high rotating speed, the dehu-

midification capacity decreases, because the reduced airflow takes more time

to cool the desiccant material and to start the dehumidification process. ' If the inlet process air absolute humidity decreases (red line), quite obviously also the outlet process air humidity decreases.

absolute humidity, with respect to the base case. ' If the inlet regeneration air temperature decreases (blue line) the relative air humidity becomes higher at constant absolute humidity. The result is a lower

dehumidification capacity, due to the reduced water mass transfer between the

desiccant and the regeneration air. ' If the regeneration air velocity decreases (green line), the moisture removal from the desiccant material is less effective. This is because the regeneration air

remains for a longer time inside the wheel, and therefore the dehumidification

capacity of the wheel is lower. ' If the inlet regeneration air absolute humidity decreases (red line), the dehu- midification capacity become higher. This effect is due to the higher water 45 6.3 Parametric analysis of performance mass transfer during the regeneration process, which leads to higher adsorp-

tion capacity. 46

of AHUs in terms of energy saving. Thus, the DW model presented in this work,

has been first validated, as shown in chapter 5, and then, it has been used for a

system-level study, that is described in this chapter. The main purposes of this

study are to show that the model is suitable for control design, and to prove the

energy saving discussed in sec. 1.5. In this chapter, we first choose a representative

operating context and the desired indoor air conditions. Then, we present a

Modelica model of a standard AHU, and we introduce a simple control system for

it. In the same operating conditions, a control system is then implemented also for

a AHU based on a DW with the aim of obtaining the same HVAC operation

required to the AHU without DW. Finally, the energy consumptions of the two

models (without and with DW) are compared through simulation.

dard AHU and AHU with DW, and we establish the desired system control perfor-

mances. The chosen air conditions, in terms of humidity and temperature, are shown in

Tab. 7.1. In the presented configurations, one third of the return air has to be

recirculated. For the AHU with DW the operation speed of the wheel has been set

at 3rev/hour. Air conditions X[g/kg] T [°C] External ambient air 17 30 Indoor return air 15 27 Desired indoor air 13 26

through simulation. Thus, we decide to compare standard AHU and AHU with DW 47 7.2 Control of a standard AHU in the same operating conditions, as said before. It is worth noting, however, that

the energy consumption is also affected by the performances of the control system.

If we desire a higher-performances control, we have to spend more in terms of energy

than if we use a control with lower performances. Therefore, we here define some

desired control performances, as shown in Tab. 7.2, and we choose a control system

that guarantees the system behavior at the boundary of such conditions. To specify

said control performances, we refer to the temperature and humidity set point step

variations, as shown in the followed section. Control performance desired value maximum overshoot [%] 25 temperature 2% settling time [sec] 60 humidity 5% settling time [sec] 1800

mixed with the recirculation air, passes through a cooling coil, in which it is cooled

down considerably. Then it has to be heated before it can enter the indoor ambient

at the desired conditions of humidity and temperature. To bring process air in such

conditions, the control variables that we can use are the cooling (qc) and heating

(qh) water mass flow rates. For the classical control system design, we need to know

the system in terms of transfer functions. Thus, we make the following simulation: 48 7.2 Control of a standard AHU we let the system reach an equilibrium state, and then we apply a step change to

each control variable, as shown in Fig. 7.2 and Fig. 7.3.

clear that the system is triangular: qh does not affect the air humidity, but only the

air temperature, while qc affects both the quantities.

functions which characterize the system using an empirical method. The result

is shown in figure Fig. 7.8, where Gcx = ''0.021839 1+60s e''5s , Ght = 14.56 1+17.25s e ''4.25s and Gct = ''71.13 1+88.5s e ''2.5s.

PI_X uses the control variable qc to control air humidity, while PI_T uses qh to

control air temperature; both qc and qh depend on valves'' opening and closure.

on the Internal Model Control (IMC) principle [15], and then with a some fine- 51 7.2 Control of a standard AHU tuning based on simulations. The results of this procedure are shown in Tab. 7.3.

This method was chosen because it seems to be a good choice in the case of model

responses characterized by a dominant time constant and a delay, as in the present

case. Controller Kp T i PI_X -5.8 60 PI_T 0.05 17

control system. Blue and red lines represent, respectively, the control set points, and

the controlled variables responses (T out and Xout). Note the temperature scale in

Fig. 7.12.

among the main control purposes in HVAC systems. Both regulations satisfy the

desired specifications listed in Tab. 7.2. Xout response is slower than T out one,

because of the desired control behavior. It is worth noting that control parameters are chosen in order to obtain a

''frequency separation' which allows the humidity control not to affect too much

the T out behavior, as shown in Fig. 7.12. 53 7.3 Control of an AHU with DW

for obvious space reasons, endowed with a control system. The external ambient

air is mixed with the recirculation air, and then it passes through the wheel process

area. In the DW, process air is dehumidified, until reaching the desired value, and

heated. Then, a cooling coil cools the air, before it enters the internal ambient. Air

from the indoor is split and used as recirculation air or as regeneration air, after

being heated. The model has the same boundary conditions and results in the same

control scenario as in the case of the standard AHU control (sec. 7.1). For control system design, we follow the same steps illustrated in sec. 7.2. The

system is still triangular, but in this case we have to use qh to control air humidity

and qc to control air temperature. Referring to Fig. 7.14 the transfer functions are:

Ghx = ''0.021839 1+325.5s e ''189.5s , G ct = ''71.72 1+70.5s e ''3.1s and G ht = 43.06 1+712.5s e ''400.5s.

higher-order dynamics, in the system. This dynamic can also be approximated with

a first order model plus delay. In this case we found consistent delays, as shown

in Ghx and Ght. The wheel performs 3 rev/hour and thus, the heat and moisture

transport is quite slow. Therefore, the system control performances are limited by

the system delay between the qh variation and the induced Xout response. However,

we can use the same control strategy, as shown for the standard AHU, and obtain

good control results. The PIs parameters used in the model of an AHU with DW

are shown in Tab. 7.4. Controller Kp T i PI_X -10 525 PI_T -0.8 60

which the set point signals vary as in the case of the standard AHU simulation in the

previous section. In Fig. 7.15, Fig. 7.16 and Fig. 7.17, blue and red lines represent

respectively, the control Set Points, and the controlled variables responses (Xout

and T out). Note the temperature scale in Fig. 7.17.

as the system has a higher-order dynamic. With a PI controller, it is not possible

to increase further the settling time performance, but with a different humidity

controller we can do so, as shown later in sec. 7.5.2. However, the slow dynamic of

Xout, even if of tens of minutes, is not a critical matter. We have to remember that

the presented control system is used for human being comfort. For such a purpose,

the control performances shown above are acceptable. 56 7.4 Comparison between standard AHU and AHU with DW

the one with DW. First, the air temperature and humidity responses are shown

together to prove the ''similar' behavior of the two control systems. Then, the

energy consumptions are compared to show the possible energy saving with the use

of a DW solution.

overshoot and settling time. In order to appreciate this comparison, simulation re-

sults shown in sec. 7.2 and sec. 7.3 have been plotted together, and shown in Fig. 7.18

and Fig. 7.19.

(green line). 57 7.4 Comparison between standard AHU and AHU with DW

in terms of performances. Thus, the energy consumptions of the two models can be

compared, as shown the next section.

DW) cases, and the control performances are similar, we can compare the energy

consumption of the two (controlled) AHU models. The heating energy, related to the heater operation, and the cooling energy, corre-

sponding to the cooler/condenser operation, are shown in Fig. 7.20 and Fig. 7.21.

These figures evidence that DW based solution (blue lines) requires less energy, es-

pecially if we consider the cooling. This result is not surprising if we consider that

the cooling energy, in the case of standard AHU, refers to the energy used by the

condenser. In that case, we require a lot of energy, because we need to pull down the

air temperature considerably. This is because we want to dehumidify the air through

condensation, thus the air has to be brought under its dew point temperature. In

the case of AHU with DW, we do not need to reduce the process air temperature so

much, because the air is dehumidified passing through the wheel, and condensation

is not required. 58 7.4 Comparison between standard AHU and AHU with DW

(w13.88 hours) are listed. The possible energy saving appear immediately clear and

it is up to 42%. 59 7.5 Control system performance Energy consumption Standard AHU AHU with DW Cooling energy [kWh] 119.73 53.5 Heating energy [kWh] 49.13 43.7 Total energy [kWh] 168.86 97.2

of the engine used for the wheel rotation. This energy is however definitely small

compared with the cooling and heating energies, thus it can be neglected in the

energy consumption comparison between the two AHU solutions.

formance. Comparing the temperature control requires with the humidity ones, in

sec. 7.1, one can note that the settling time for humidity control is very long. In the

case of an AHU with DW, it is clear that such a time is influenced by the higher

order dynamic introduced by the wheel, as said in sec. 7.3. In a standard AHU,

conversely, that effect does not appear, thus it seems reasonable that the control

system can reach higher performances in terms of settling time. As said in sec. 7.1,

the aim of this chapter is to perform an energy comparison between the two AHU

configurations. Thus, we need the same control system performances for both the

configurations. To obtain that, we chose a set of control specifications (in terms of

overshoot and settling time of the set point response) feasible for both the standard

AHU and that with the DW, and we tuned the controls of the two in such a way

to best approach those specifications. Of course this means that no optimality was

pursued for the two units individually, but rather a simulation setting was created

where they are required to ''do the same (reasonable) things', and the two energy

consumptions for that operation are compared. In this section, we present a par-

ticular choice of PI_X parameters for the standard AHU that guarantees better

performances and we show the energy consumption in that case. The second topic of this section deals with the AHU with DW control. As said in

sec. 7.3, with a PI controller, it is not possible to make faster the humidity response,

but with a different controller, we can do something about it. Now, we want to

present a simple PID controller that exploits this possibility. 60 7.5 Control system performance

PI_X parameters are different from the previous case, as shown in Tab. 7.6. Controller Kp T i old PI_X -5.8 60 new PI_X -80 60

same signal in the case of standard AHU control illustrated in sec. 7.2. It is clear that

with a different humidity controller, we can reduce the settling time significantly.

greater coupling of the system, as shown in Fig. 7.23. This is clearly an undesired

effect, but the results obtained herein seems to suggest that it can be considered a

small drawback compared to the performance improvement on the humidity control.

The matter surely deserves a deeper analysis, that is however deferred to future

research. 61 7.5 Control system performance

desire a higher-performance control, we generally have to spend more in terms of

energy. Fig. 7.24 and Fig. 7.25 show that the system configuration with the new

PI_X parameters, is a little more energy demanding compared with that with the

old PI_X parameters. The energy consumption in the reference period is shown in

Tab. 7.7 for both cases. Based on the obtained results, we can confirm the energy

saving of the DW solution as in the case of standard AHU with control performances

closer to the real operation.

one devoted to the temperature control, and the other to the humidity control.

Such an approach proves suitable for the considered control purpose, but for a more

general viewpoint, the system performances could be improved with different control

schemes. It is not within the scope of the present work to analyze advanced control

strategies, nevertheless we spend just a few words on a different humidity controller

choice. We focus on the humidity control because it seems to be the one with the

most critical behavior. 63 7.5 Control system performance

behavior with a first order plus delay model. This approach is simple and leads to

a quite good controller choice. However, if we analyze more in detail the system

humidity response (Fig. 7.26), this seems to be more like a response of a system with

higher-order dynamics. Thus, we can try to approximate it with a second order

dynamic, in order to use a PID controller. The use of derivative action allows to

reduce the settling time; further, a PID controller is easily implemented in industrial

applications. The PID parameters chosen for the new humidity controller, are listed

in Tab. 7.8 (N represents the derivative filter ratio). Controller Kp T i T d N PID_X -35 550 130 200

tling time and overshoot reduction as shown in Tab. 7.9. 64 7.5 Control system performance

control system performances. However, the difference between the presented control

scheme and that with two PI controllers, in terms of energy consumptions, is not

so relevant, thus the obtained response improvements could be worth the expense,

although dealing with such tradeoffs in a methodologically sound manner is outside

of our scope. 65

dard AHU and an AHU with DW was presented. Thus, we proved the (considerable)

energy saving that can be obtained with the use of a DW. In the control configu-

ration, shown in the previous chapter, we used the heater water flow rate (qh) to

control process air humidity and the cooler water flow rate (qc) to control the air

temperature. The DW velocity (') was considered as a parameter. In this chapter,

we investigate the possible use of ' as a control variable.

is clear that the energy saving is more relevant if we focus on the cooling energy. As

said in sec. 7.4.2, the cooling energy in the case of AHU with DW, is related to the

cooler operation. Thus, in the AHU, the cooler is the critical element if we consider

the energy consumption. Further, it is worth noting that changing the wheel velocity

affects the energy consumption of the AHU, but without a relevant effort in terms

of wheel engine energy. For these reasons, the question appears legitimate, whether

or not the use of ' as a control variable, instead of qc, could be a good choice.

Thus, the idea is to control process air temperature and humidity with ' and qh.

Incidentally, with this new control configuration, the use of the cooling coil is no

more necessary. First, we decided to choose some inlet process and regeneration air conditions and to

investigate how the wheel velocity and inlet regeneration air temperature (T reg, in)

affect the process air outlet conditions. Note that T reg, in in the AHU scheme

depends on qh. We performed a set of simulations with the wheel model used in chapter 5, and the

inlet conditions and parameters in Tab. 8.1. We choose a range of variations for '

and T reg, in and a step of variation for each parameter, as shown in Tab. 8.2. 66 8.1 Investigation on the use of the wheel velocity as a control variable simulation parameters Xpro, in[g/kg] Xreg, in[g/kg] T pro, in[°C] Apro

Areg [m 2/m2] sim 16.0 15.0 30.0 1

The curves represent the simulations with the same ', and the dots on each curve

represent different values of T reg, in (' increases from left to right and T reg, in

increases from top to bottom along each curve).

tions of ' and T reg, in. Recalling that T pro, in = 30°C, it is clear that, if we want to

dehumidify, the process air temperature always increases passing through the wheel.

Thus, the proposed control scheme is not useful for our purposes (dehumidify and

cool down the air). Taking into account the wheel operation, it is possible to im-

plement such a control scheme only if we want to dehumidify and heat the process 67 8.2 Wheel velocity control air, thus when the external ambient air is ''cold' and with a consistent moisture

amount. This situation is most typical in winter, however could happen also in

summer operation, e.g. in the nigh time and in the particular, but relevant, case

of storms. In the situation just mentioned, the external ambient air temperature

is lower than the required air temperature, and we need to dehumidify. A heater

coluld be placed after the DW, and used to increase the process air temperature

once it exits the wheel, and the cooler is switched off. Here, we propose to use the

wheel velocity to increase the air temperature without using an additional heater.

with the same control scheme used in Fig. 7.13. We maintained the same indoor

ambient air conditions, air temperature and humidity set points (Tab. 8.3), PIs''

parameters and wheel velocity (3rev/h) as in sec. 7.3, while the external ambient air

temperature and humidity follow the profiles shown in Fig. 8.2 and Fig. 8.3. Note

that we considered a day long cycle for both air temperature and humidity. We

decided to take into account the second day because of the startup effects. airflow temperature ['C] humidity [g/kg] indoor air 27 15 set point 26 13

point during almost half a period, while the humidity is always higher than the set

point. As said in sec. 8.1, the air temperature always increases passing through the

wheel. It is clear that if, as in the presented scenario, the process air temperature

does not grow enough to excede the set point temperature, we need to increase

T pro, out. In Fig. 8.4, we show the simulation results: during a relevant period, the

desired air temperature could not be reached since T pro, out is lower than the set

point. Therefore, to reach the desire temperature we should heat the air exits the

wheel, but in the presented AHU configuration we have only a cooler after the DW

and thus we can only switch it off.

higher than the required air temperature. For such a purpose, we added a PI controller that uses T pro, out as controlled variable, the wheel velocity as control

variable and the same set point of the PI_T controller. A small offset can be added

to the set point in order to maintain T pro, out a little bit higher than the required

temperature. The control scheme is shown in Fig. 8.5.

the desired temperature set point. Thus, we introduced a minimum saturation level

in the control signal. This minimum level must be chosen properly, and a possible

method for that choice will be discussed in the next chapter. Obviously, even a

maximum saturation value has to be set in order to maintain the wheel in the range

of operation. We repeat the previous simulation with the outlined control scheme.

In Tab. 8.4 we show the PI_' parameters. Controller kp T i ' min[rev/h] ' max [rev/h] of f set PI_' 0.1 200 2 5 0

the set point for the entire simulation period. Thus, we can ensure the desire air

temperature without the need of a post heating. 70 8.2 Wheel velocity control

desired set point, as shown in Fig. 8.7. When the velocity control is not necessary,

the wheel speed is set at the low saturation value. The choice of this value is very

important because the wheel operation depends on it for the most of time. Thus,

this velocity has to be chosen to guarantee the best performances for the system.

The considerations developed in the next chapter for the choice of the best wheel

velocity, can be used also to determine this minimum value.

process air, the control scheme adopted in chapter 7 is required. However, the choice

of the wheel velocity could be made with different approaches. In this chapter, to

give just an example of how the idea can be exploited, we pursue the energy saving

policy.

element is the cooler. Quite intuitively, the energy required by the cooler depends

on its inlet air temperature. The higher the inlet air temperature is, the greater

the energy consumption will be. The considered temperature depends on the wheel

operation and corresponds to T pro, out. Thus, if we succeed in reaching the desired

Xpro, out with the minimum T pro, out, the energy consumption of the cooler will

be the minimum possible. 72 9.1 Effect of the wheel velocity on the energy consumption

we choose the lowest wheel velocity that permits to reach the desired air moisture

content, the outlet process air temperature reaches the minimum value. The same figure also shows another interesting result. The same value of Xpro, out

is reached for a different value of T reg, in depending on the wheel velocity. For

example, if we consider the curves with ' = 0.5 rev/h and ' = 1 rev/h the same

value of Xpro, out is reached for a greater T reg, in in the case of ' = 0.5 rev/h.

Thus, the choice of the minimum velocity results in the need of higher regeneration

air temperature to reach the desired humidity value for the process air. The air

entering the heater is considered at constant temperature, and obviously, to bring

the regeneration air at a high temperature, the heater requires more energy. Therefore, we have to consider two opposite effects that occur when reducing the

wheel velocity: the decrease of cooling energy demand, and the increase of heating

energy one. In order to better visualize this effect, we show some simulation results dealing with

the energy demand with respect to different wheel velocities. Fig. 9.2 shows the total

energy consumption, while Fig. 9.3 separately shows the heating and cooling energy

demand for the same simulations. 73 9.1 Effect of the wheel velocity on the energy consumption

trend, while the heating energy has a minimum value near ' = 3 rev/h. For 74 9.2 Total energy demand minimization ' < 3 rev/h cooling energy increases while heating energy decreases. Thus, the

total energy consumption, for low wheel velocity, results more or less constant. For

higher wheel speed, conversely, both the energy consumptions increase with ', thus

the total energy consumption increases considerably. Summarising, in any case, the previous considerations show that we can minimize

the total energy consumption with an appropriate choose of the wheel velocity.

the DW velocity that corresponds on the minimum of the AHU energy consumption.

Note that the model includes the control scheme, and thus the system is in closed

loop. We decided here too, to define a base case, and two types of simulations. In the

first type (type I), we took into account the external air conditions variations due

to climate changes. Thus, we maintained the indoor air conditions and varied the

external ambient air temperature and humidity. In the second type (type II) of

simulations, we considered changes in the desired internal air conditions. Thus,

we maintained as constant the external air conditions and varied the humidity and

temperature set points. The temperature and humidity values for external ambient air and controllers'' set

points, used in the base case and type I simulations, are shown in Tab. 9.1. Simulation Xamb [g/kg] T amb [°C] Xset point [g/kg] T set point [°C] base case 17.0 30 13.0 26 simA

the total energy consumption and in the second one, we show a comparison between

cooling and heating energy. Fig. 9.4 and Fig. 9.6 show the total energy consumptions

depending on the wheel velocity. The first figure shows the simulations in which

a change in the ambient humidity is performed, while the second one deals with

simulations with a different ambient temperature, referring to the base case. 75 9.2 Total energy demand minimization

dehumidification demand. If we desire a higher dehumidification, the trend of the 76 9.2 Total energy demand minimization heating energy is more accentuated. Thus, the total energy consumption exhibits a

minimum, that depends on the heating energy curve.

simulation, but for greater value of ' we have a relevant increase in the consumption. Therefore, a choice of DW velocity under 2 rev/h seems to be the best for all possible

ambient conditions changes around the considered scenario, that is represented by

the base case. Now, we analyze the second type of possible operating conditions changes. The

temperature and humidity set points values, used in the base case and type II

simulations, are shown in Tab. 9.2. Simulation Xamb [g/kg] T amb [°C] Xset point [g/kg] T set point [°C] base case 17.0 30 13.0 26 simE 17.0 30

said for type I simulations, for ' < 2 rev/h we have more or less the same energy

consumption for each simulation, but for greater values of ' we have a relevant

increase in the energy demand, especially if we desire a greater dehumidification

(simE). It is worth noting that if we desire a lower or higher air temperature, the

consumption curves (simG and simH) differ from the base case only by a (practically

constant) offset. This is because the temperature control does not involve directly

the wheel operation, hence the increase of energy demand is due only to the cooling

energy. Thus, in Fig. 9.11 we can see that the heating energy consumption is almost

the same for the base case and for simulations G and H. 78 9.2 Total energy demand minimization

to minimize the total energy consumption. 80 9.3 Cooling energy demand minimization As said before, the minimization of the total energy is a way to take into account

both the cooling and the heating consumptions depending on the wheel velocity.

Another interesting choice could be the minimization of the cooling energy only in

some operative conditions. This choice could be better, referring to the total energy

minimization, if e.g. we can use the heat of recovery from other applications to

regenerate the wheel. In the following section, we analyze the wheel velocity choice

considering only the cooling energy.

in the previous section, concentrating however on the cooling energy. If we consider

only the cooling energy demand, it is clear that the choice of the DW velocity as the

minimum as possible is the right one. This assumption, mentioned before, when we

deal with T pro, out minimization, is supported by the simulations results. If we use,

for the wheel regeneration, a low temperature heat source, we have also to consider

the feasibility of the process because reducing the wheel velocity, the regeneration

air temperature demand increases.

different choice of the wheel velocity affects the control system performances. To

this end, we decided to make a simulation with ' = 0.5 rev/h. We choose such

a value because it is the one, in the optimal operating range (' < 2 rev/h), that

is farthest from the value we used in sec. 7.3 (' = 3 rev/h). We compare the

system performances with ' = 3 rev/h and ' = 0.5 rev/h in terms of air humidity,

Fig. 9.16, and and temperature responses, Fig. 9.17. 83 9.4 System performances with different choice of wheel velocity

line).

(green line). We can notice that the control system behavior is quite similar in both the cases,

and the desired system performances are preserved. Thus, the use of a quite different

wheel velocity affects the energy consumption without changing the system behavior

significantly. In Fig. 9.18 and Fig. 9.19, the cooling and heating energy consumptions

for the different DW velocities are shown. In Tab. 9.3, the energy consumptions

during the entire simulation period of 50000 sec (w13.88 hours) are listed. In extreme

synthesis, although the matter requires further research as already stated, we can

say that the total energy saving is confirmed by the simulations. 84 9.4 System performances with different choice of wheel velocity

addressed. The presented work is part of a long-term research on energy efficiency

in HVAC systems, and in that context, its primary goal is to investigate the dynamic

behavior of desiccant wheels, particularly in a view to using such devices for energy

efficiency improvements in air handling units. To this end, first the theory of operation of desiccant wheels was analyzed (Capitolo 2),

based on first-principle considerations and on a convenient review of related work,

reported in Capitolo 3. A specific character of the reported analysis is its focus on

dynamic models oriented to system-level studies, that apparently have to consider

also the control system. As a result, in Capitolo 4, a dynamic model of a DW was

proposed, and then validated in chapter 5. The presented model has an intermedi-

ate complexity with respect to the two main alternatives in the literature, and thus

- as illustrated - it is suitable for our purposes. One of the main advantages of using

the presented model is that the results can be easily interpreted, because of the

physical meaning of the parameters. Further, also the choice of how to model the

wheel motion brings some relevant advantages both in the efficiency of the model

and the understanding of the simulation results. Thus, the model is suitable for both

system-level studies and control purposes, and can also be integrated with others

from different physical domain thanks to its OOM character. In chapter 7, the focus was moved from the DW model to a wider context in which

the model can be integrated, namely - to stick to the dissertation scope - that of an

AHU for an HVAC system. We proposed some control schemes for both a standard

AHU and an AHU with DW and we performed an energy consumption comparison,

thus confirming the saving yielded by use of the DW based solution. In chapter 8 we investigated some alternative (with respect to the traditional so-

lutions) control approaches, specifically aimed - among the various and often con-

flicting possible objectives - at energy efficiency. The possibility and opportunity

of using the DW velocity as a control variable was synthetically analyzed, and we

also proposed a temperature control strategy using the DW velocity. Another ap-

proach, dealing with the wheel velocity, is shown in chapter 9 in which we proposed

a method, devoted to minimize the energy consumption of the AHU, by choosing

an appropriate value for the DW velocity. Future research will be devoted to investigate case studies of higher complexity and

different control approaches in a rigorous manner, with the aim to better exploit the

model advantages in the context of energy efficiency and design optimization. 86

control loops. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.9 Control scheme for humidity and temperature control of an AHU with

a DW. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.10 Energy consumption in the two cases: standard and DW layout. . . . 13 2.1 Desiccant wheel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.2 Adsorption isotherms. . . . . . . . . . . . . . . . . . . . . . . . . . . 16 2.3 Adsorption isotherm hysteresis. . . . . . . . . . . . . . . . . . . . . . 17 2.4 Air outlet water content for different desiccant wheel rotating speeds. 18 2.5 Pennington cycle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.6 Recirculation cycle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.7 Staged regeneration cycle. . . . . . . . . . . . . . . . . . . . . . . . . 21 3.1 Wheel motion representation for Casas et all. . . . . . . . . . . . . . 24 3.2 Wheel motion representation for Joos et al.. . . . . . . . . . . . . . . 25 3.3 Wheel motion from a thermo-hydraulic viewpoint. . . . . . . . . . . . 25 4.1 Wheel structure and spatial discretisation. . . . . . . . . . . . . . . . 27 4.2 Single channel structure. . . . . . . . . . . . . . . . . . . . . . . . . . 29 4.3 Desiccant flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 4.4 Air, water and desiccant mass flows for the control volume . . . . . . 33 5.1 Process and regeneration air flow scheme. . . . . . . . . . . . . . . . . 36 5.2 Comparison between experimental and simulated process outlet hu-

midity for different working conditions (wheel rotation speed). . . . . 37 6.1 Spatial discretisation of the wheel. . . . . . . . . . . . . . . . . . . . . 40 6.2 Air temperature.Spatial distribution. . . . . . . . . . . . . . . . . . . 41 6.3 Desiccant material temperature. Spatial distribution. . . . . . . . . . 41 87 List of Figures 6.4 Water content in the air. Spatial distribution. . . . . . . . . . . . . . 42 6.5 Water content in the desiccant material. Spatial distribution. . . . . . 43 6.6 Outlet process air humidity versus rotating speed: influence of process

air conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 6.7 Outlet process air humidity versus rotating speed: influence of regen-

eration air conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . 45 7.1 Standard AHU model in Modelica . . . . . . . . . . . . . . . . . . . . 48 7.2 Step on heating water mass flow rate (qh). . . . . . . . . . . . . . . . 49 7.3 Step on cooling water mass flow rate (qc). . . . . . . . . . . . . . . . 49 7.4 Air humidity (Xout) response. . . . . . . . . . . . . . . . . . . . . . 49 7.5 Air temperature (Tout) response. . . . . . . . . . . . . . . . . . . . . 50 7.6 Air humidity (Xout) response. . . . . . . . . . . . . . . . . . . . . . . 50 7.7 Air temperature (Tout) response. . . . . . . . . . . . . . . . . . . . . 50 7.8 AHU block diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 7.9 AHU control scheme. . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 7.10 Step change in temperature set point and air temperature response. . 52

7.11 Step change in humidity set point and air humidity response. . . . . . 53

7.12 Air temperature variation due to the step of humidity set point. . . 53 7.13 AHU with DW model in Modelica. . . . . . . . . . . . . . . . . . . . 54

7.14 AHU with DW block diagram. . . . . . . . . . . . . . . . . . . . . . . 54

7.15 Step change in temperature set point and air temperature response. . 55

7.16 Step change in humidity set point and air humidity response. . . . . . 56

7.17 Air temperature variation due to the step of humidity set point. . . . 56

7.18 Step change in temperature set point (blue line) and air temperature response. Comparison between standard AHU (red line) and AHU

with DW (green line). . . . . . . . . . . . . . . . . . . . . . . . . . . 57 7.19 Step change in humidity set point and air humidity response. Com- parison between standard AHU (red line) and AHU with DW (green

line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 7.20 Heating energy consumption. Comparison between standard AHU (red line) and AHU with DW (blue line). . . . . . . . . . . . . . . . . 59 7.21 Cooling energy consumption. Comparison between standard AHU (red line) and AHU with DW (blue line). . . . . . . . . . . . . . . . . 59 7.22 Step change in humidity set point (blue line) and air humidity re- sponse. Comparison between new PI_X (red line) and old PI_X

(green line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 7.23 Air temperature variation due to the step of humidity set point. Com- parison between new PI_X (red line) and old PI_X (green line). . . . 62 7.24 Heating energy consumption. Comparison between standard AHU with the old PI_X (blue line) and the one with the new PI_X (red

line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 88 List of Figures 7.25 Cooling energy consumption. Comparison between standard AHU with the old PI_X (blue line) and the one with the new PI_X (red

line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 7.26 Air humidity (Xout) response. . . . . . . . . . . . . . . . . . . . . . . 64

7.27 Step change in humidity set point (blue line) and air humidity re- sponse. Comparison between new PID_X (green line) and old PI_X

(red line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 8.1 Operation range. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 8.2 External ambient air temperature profile (blue line) and temperature

set point (red line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 8.3 External ambient air humidity profile (blue line) and humidity set

point (red line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 8.4 DW outlet process air temperature (blue line) and air temperature

entering the room (red line). . . . . . . . . . . . . . . . . . . . . . . 69 8.5 AHU with DW model. Wheel velocity control scheme. . . . . . . . . 70 8.6 DW outlet process air temperature (blue line) and air temperature

entering the room (red line). . . . . . . . . . . . . . . . . . . . . . . . 71 8.7 Wheel velocity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 9.1 Operation range. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 9.2 Total energy consumption. . . . . . . . . . . . . . . . . . . . . . . . . 74 9.3 Heating and cooling energy consumption. . . . . . . . . . . . . . . . . 74 9.4 Total energy consumptions for simulations A and B. . . . . . . . . . . 76 9.5 Cooling and heating energy consumptions for simulations A and B. . 76 9.6 Total energy consumptions for simulations C and D. . . . . . . . . . . 77 9.7 Cooling and heating energy consumptions for simulations C and D. . 77 9.8 Total energy consumptions for simulations E and F. . . . . . . . . . . 79 9.9 Cooling and heating energy consumptions for simulations E and F. . 79 9.10 Total energy consumptions for simulations G and H. . . . . . . . . . . 80

9.11 Cooling and heating energy consumptions for simulations G and H. . 80

9.12 Cooling energy consumptions for simulations A and B. . . . . . . . . 81

9.13 Cooling energy consumptions for simulations C and D. . . . . . . . . 82

9.14 Cooling energy consumptions for simulations E and F. . . . . . . . . 82 9.15 Cooling energy consumptions for simulations G and H. . . . . . . . . 83

9.16 Step change in humidity set point and air humidity response. Com- parison between AHU with DW: ' = 3 rev/h (red line) and ' = 0.5

rev/h (green line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 9.17 Step change in temperature set point and air temperature response. Comparison between AHU with DW: ' = 3 rev/h (red line) and

' = 0.5 rev/h (green line). . . . . . . . . . . . . . . . . . . . . . . . . 84 9.18 Cooling energy consumption. Comparison between AHU with DW: ' = 3 rev/h (blue line) and ' = 0.5 rev/h (red line). . . . . . . . . . 85 89 List of Figures 9.19 Heating energy consumption. Comparison between AHU with DW: ' = 3 rev/h (blue line) and ' = 0.5 rev/h (red line). . . . . . . . . . 85 90

and experimental data (exp). . . . . . . . . . . . . . . . . . . . . . . 37 6.1 Base case inlet air flows conditions. . . . . . . . . . . . . . . . . . . . 39 7.1 Air conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 7.2 Desired control performances. . . . . . . . . . . . . . . . . . . . . . . 48 7.3 PIs parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 7.4 PIs parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 7.5 Energy consumptions. Comparison between standard AHU and AHU

with DW. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 7.6 PI_X parameters. Old and new solution. . . . . . . . . . . . . . . . . 61 7.7 Energy consumptions. Comparison between standard AHU with old

PI_X parameter and AHU with new PI_X parameter. . . . . . . . . 63 7.8 PID_X parameter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 7.9 AHU with DW humidity performance. Comparison between new PID_X and old PI_X. . . . . . . . . . . . . . . . . . . . . . . . . . . 65 8.1 Setup parameters for simulations. . . . . . . . . . . . . . . . . . . . 67 8.2 Range of variation and step for wheel velocity and regeneration air

temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 8.3 Simulation parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . 68 8.4 PI_' parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 9.1 Simulations parameters (type I). . . . . . . . . . . . . . . . . . . . . . 75 9.2 Simulations parameters (type II). . . . . . . . . . . . . . . . . . . . . 78 9.3 Energy consumptions. Comparison between AHUs with DW. . . . . . 85 91

engineering of Japan, 26:530''535, 1993. [3] A.Kodama, T.Hirose M.Goto, and T.Kuma. temperature profile and optimal rotational speed of a honeycomb rotor adsorber operated with terma swing.

Journal of Chemical engineering of Japan, 27:644''649, 1994. [4] A.Kodama, T.Hirose M.Goto, and T.Kuma. Performance evaluations for a thermal swing honeycomb rotor using a humidity chart. Journal of Chemical

engineering of Japan, 28:19''24, 1995. [5] M.Goto A.Kodama, T.Hirayama, T.Hirose, and R.Critoph. The use of psychro- metric charts for the potimisation of a thermal swing desiccant wheel. Applied

Thermal Energy, 21(16):1657''1674, 2001. [6] H.S. Al-Sharqawi and N.Lior. Effect of flow-duct geometry on solid desiccant dehumidification. Ind. Eng. Chem. Res., 47:1569''1585, 2008. [7] S. De Antonellis, C.M.Joppolo, and L.Molinaroli. Simulation, performance analysis and optimization of desiccant wheels. Energy and Buildings, 42:1386''

1393, 2010. [8] B.O.Glav. Air conditioning apparatus. Patent No.3,251,402, 1966. [9] Y.Li D.La, Y.J. Dai, R.Z.Wang, and T.S.Ge. Technical development of rotary desiccant dehumidification and air conditioning: A review. Renewable and Sustainable Energy Revuews, 14:130''147, 2010. [10] D.Murayama, K.Mitsumoto, and S.Yamamori Y.Takagi, Y.Iino. Smart grid ready bems adopting model-based hvac control for energy saving. [11] J.J. Jurinak. Open-cycle solid desiccant cooling. Component models and system simulation. PhD thesis, University of Wisconsin, Madison, 1982. [12] F.Butera M.Beccali, R.Guanella, and R.S.Adhikari. Simplified models for the performanche evaluation of desiccant wheel dehumidification. International Journal of Energy Res, 27(1):17''29, 2003. [13] M.Bonvini and A.Leva. Modelling and control of desiccant wheel for energy- efficient air handling units. 92 Bibliography [14] N.A.Pennington. Humidity changer for air conditioning. Patent No.2,700,537, 1955. [15] I.Postlethwaite S.Skogestad. Multivariable Feedback Control - Analysis and De- sign. Wiley, 2005. [16] Y.Li T.S.Ge, R.Z.Wang, and Y.J. Dai. A review of the mathematical models for predicting rotary desiccant wheel. Renewable and Sustainable Energy Revuews,

12:1485''1528, 2008. [17] W.Casas, K.Prolb, and G.Schmitz. Modeling of desiccant assisted air con- ditioning systems. Proceedings of the 4th International Modelica Conference,

2005. [18] A.L. London W.M. Kays. Compact Heat-Exchangers. third edition, 1984. 93

Grassi Mobil™ - Formulati per fornire elevate prestazioni anche in condizioni operative estreme

© Eiom - All rights Reserved P.IVA 00850640186